Re: [PATCH 03/10] KVM: SVM: Terminate the VM if a SEV-ES+ guest is run with an invalid VMSA

[Sean Christopherson <seanjc@google.com>]

On Mon, Feb 24, 2025, Sean Christopherson wrote:
> On Mon, Feb 24, 2025, Tom Lendacky wrote:
> > On 2/24/25 16:55, Sean Christopherson wrote:
> > > On Mon, Feb 24, 2025, Tom Lendacky wrote:
> > >> On 2/18/25 19:26, Sean Christopherson wrote:
> > >>> -void pre_sev_run(struct vcpu_svm *svm, int cpu)
> > >>> +int pre_sev_run(struct vcpu_svm *svm, int cpu)
> > >>>  {
> > >>>  	struct svm_cpu_data *sd = per_cpu_ptr(&svm_data, cpu);
> > >>> -	unsigned int asid = sev_get_asid(svm->vcpu.kvm);
> > >>> +	struct kvm *kvm = svm->vcpu.kvm;
> > >>> +	unsigned int asid = sev_get_asid(kvm);
> > >>> +
> > >>> +	/*
> > >>> +	 * Terminate the VM if userspace attempts to run the vCPU with an
> > >>> +	 * invalid VMSA, e.g. if userspace forces the vCPU to be RUNNABLE after
> > >>> +	 * an SNP AP Destroy event.
> > >>> +	 */
> > >>> +	if (sev_es_guest(kvm) && !VALID_PAGE(svm->vmcb->control.vmsa_pa)) {
> > >>> +		kvm_vm_dead(kvm);
> > >>> +		return -EIO;
> > >>> +	}
> > >>
> > >> If a VMRUN is performed with the vmsa_pa value set to INVALID_PAGE, the
> > >> VMRUN will fail and KVM will dump the VMCB and exit back to userspace
> > > 
> > > I haven't tested, but based on what the APM says, I'm pretty sure this would crash
> > > the host due to a #GP on VMRUN, i.e. due to the resulting kvm_spurious_fault().
> > > 
> > >   IF (rAX contains an unsupported physical address)
> > >     EXCEPTION [#GP]
> > 
> > Well that's for the VMCB, the VMSA is pointed to by the VMCB and results
> > in a VMEXIT code of -1 if you don't supply a proper page-aligned,
> > physical address.
> 
> Ah, good to know (and somewhat of a relief :-) ).

If anyone else was wondering, the behavior is described in the "VMRUN Page Checks"
table, which "Table 15-4" in the March 2024 version of the APM.

FWIW, knowing that VMRUN is supposed to handle this scenario does make me somewhat
tempted to skip this patch entirely.  But I still don't like the idea of doing
VMRUN with a known bad address.